Geospatial modelling of post-cyclone Shaheen recovery using nighttime light data and MGWR

Tropical cyclones are a highly destructive natural hazard that can cause extensive damage to as-sets and loss of life. This is especially true for the many coastal cities and communities that lie in their paths. Despite their significance globally, research on post-cyclone recovery rates has gener-ally been qualitative and, crucially, has lacked spatial definition. Here, we used freely available satellite nighttime light data to model spatially the rate of post-cyclone recovery and selected sev-eral spatial covariates (socioeconomic, environmental and topographical factors) to explain the rate of recovery. We fitted three types of regression model to characterize the relationship be-tween rate of recovery and the selected covariates; one global model (linear regression) and two local models (geographically weighted regression, GWR, and multiscale geographically weighted regression, MGWR). Despite the rate of recovery being a challenging variable to predict, the two local models explained 42% (GWR) and 51% (MGWR) of the variation, compared to the global linear model which explained only 13% of the variation. Importantly, the local models revealed which covariates were explanatory at which places; information that could be crucial to policy -makers and local decision-makers in relation to disaster preparedness and recovery planning.